(1) Environmental heterogeneity has two components: the overall variance of sites within an area, and the way in which the variance between sites changes with the distance separating them. It can be studied indirectly through the measurement of physical attributes of the environment, or directly through the response of organisms to the environment. Studies using direct bioassays are of two kinds: implant studies, in which plants are grown in the natural environment, and explant studies, in which plants are grown in samples of the environment, such as soil cores, under controlled conditions. Explant studies give conservative estimates of heterogeneity. (2) This paper describes explant trials within a 50-m x 50-m area of undisturbed Fagus grandifolia-Acer saccharum forest situated south-east of Montreal, Quebec. Three sampling schemes were used: random sampling, to yield an unbiased estimate of variation; nested sampling, to yield preliminary estimates of spatial structure (i.e. correlation of nearby sites); and grid sampling, to yield detailed information on structure at spatial scales of 50 m, 10 m and 1 m. Two series of trials were conducted, a bioassay using Arabidopsis thaliana and another using Hordeum vulgare. Two replicate soil cores for each of two genotypes of each species were collected at each site in each trial. Plants were sown into these cores in the glasshouse, and several characters scored on each, including dry mass at harvest, which is closely related to reproductive success. (3) Random sampling gave substantial and significant estimates of environmental variance for most characters, including all measures of dry mass in both species. The environmental correlation for dry mass scores was about 0.25-0.40. Nested sampling showed that variance generally decreased from scales of 10 m through 1 m to 0.1 m. Grid sampling showed that the variance among sites increases with the distance separating them. The regressions of log variance on log distance were linear, with slopes, for the dry mass scores, of about 0.15-0.25. Environmental structure was detected in this way even within 1-m quadrats among sites 0.1 m apart. (4) The correlation between sites decreases with increasing separation on all spatial scales examined. Moreover, the correlation structure was independent of spatial scale: correlation declined at the same rate at all scales, when distance is expressed as the absolute distance divided by the grid dimension. Environments are about equally complex at all spatial scales.
